1. Field
The following description relates to an Augmented Reality (AR) apparatus and method for recognizing objects.
2. Discussion of the Background
Augmented Reality (AR) is a computer graphic technique of synthesizing a virtual object or virtual information with a real environment such that the virtual object or virtual information appears like a real object or real information that may exist in the real world environment.
AR is characterized in synthesizing virtual objects based on the real world to provide additional information that cannot be easily obtained from the real world, unlike existing Virtual Reality (VR) that targets only virtual spaces and virtual objects. Based on this characteristic of AR, the AR can be applied to various real environments unlike the existing VR that has been applied to limited fields such as games. AR's characteristic has allowed the AR technology to come into the spotlight as a next-generation display technique suitable for a ubiquitous environment.
For example, AR may be implemented as a method of overlapping information on an image of the real world. If a tourist points at a specific direction with a camera of a mobile phone in which a GPS sensor is installed, AR data related to the various restaurants or stores on sale located on a street within a real world image of the street captured by the mobile phone may be displayed.
In order to provide such AR data, objects that exist in the real world may be recognized. That is, the ability to recognize stores or specific items from which AR data has to be obtained is a factor in an AR technology.
Conventional recognition techniques could recognize pre-designated markers, but there were difficulties in ensuring a high recognition rate using a makerless-based recognition technique such as detection of edges and boundaries. In order to provide a more effective AR service, improvement of such a markerless-based recognition technique may be developed.
Furthermore, the conventional recognition techniques could not easily detect objects of interest from image information in which multiple objects are included. Also, the conventional recognition techniques may recognize objects in which a user is not interested and provide metadata related to the unintended objects, which may undesirably consume processor or time resources. Additionally, a large amount of time may be undesirably consumed to find a desired object among similar prestored objects to identify the recognized object.